Building AI World Models to
Make Cancer Predictable

We are redefining how causality is applied in oncology, developing AI systems that can reason about the true biological effects of treatment rather than just recognizing patterns in data. By constructing multi-scale causal models that trace how interventions alter disease across molecular, cellular, and tissue levels, our research aims to reveal why certain therapies work, for whom, and under what conditions. 

To learn from the raw complexity of medical data, our AI team is pioneering foundation models that uncover hidden biological structure within pathology. Through self-supervised and multimodal learning tailored to medicine, our models extract rich representations that generalize across diseases and clinical contexts, enabling more powerful, scalable AI systems that advance how we understand and treat disease.

In oncology, risk unfolds over time. Our AI team is advancing survival analysis with AI models that integrate pathology, clinical, and molecular data to predict not just whether a patient will recur or survive, but when. Unlike traditional approaches, our models handle complex, multi-institutional data to deliver precise, time-based prognosis, forming the foundation for personalized treatment planning and causal modeling of therapy effects.